Brimacombe Lecture

Research on Sustainable Steelmaking

R. J. FruehanCenter for Iron and Steelmaking Research

Materials Science and Engineering DepartmentCarnegie Mellon University

Pittsburgh PATel: 412.268.2683Fax: 412.268.7247

e-mail: fruehan@cmu.edu

ABSTRACT

The international steel community is faced with the challenge to develop processes which will make steel production more sustainablein the future. Specifically processes which produce less CO2, less net waste materials and emissions, and consume less energy arerequired. This paper outlines where energy consumption and CO2 emissions are high and can be reduced. Reductions can beachieved by incremental improvements to existing processes or by “break-through innovative process”; both strategies are examined.Since most of the energy consumption and CO2 generation occurs in ironmaking, research in this area is emphasized. Research oncontrolling the cohesive zone in the blast furnace, improving the final stages of reduction in direct reduction processes, the use ofbiomass and other innovative processes for ironmaking are reviewed. In oxygen steelmaking, improved post combustion to allow formore scrap melting is examined. Post combustion and slag foaming in the EAF in order to reduce energy is reviewed.

INTRODUCTION

It is not only a great honor to give the Brimacombe Lecture but also a large responsibility. It is a responsibility to be sure that theLecture is on a significant subject, makes our community sensitive to long term critical issues and is consistent with ProfessorBrimacombe’s vision of what our industry should be doing in the future. With this goal, I have chosen to discuss sustainablesteelmaking and examples of research aimed at this goal.

Current Status

Before discussing specific research projects, the overall goal of sustainability will be reviewed. Sustainable Development has beendefined as:

“Development that meets the needs of the present without compromisingthe ability of future generations to meet their needs”

Specifically for sustainable steelmaking, the goals include:

• Conserving resources such as ore, coal, etc.• Reducing the amount of CO2 emitted.• Reducing other gaseous emissions such as SO2, Nox, etc.• Reduce materials to landfills.• Reduce the amount of hazardous wastes.

Steel production remains high and continues to grow. Steel production occurs in virtually every major country with about 50%produced in five countries. Much of the future increases will be in developing countries such as China, India and those in EasternEurope. Energy consumption is about 18-20 GJ per tonne. The steel industry in the USA consumes about 18-20 GJ/t or about 2-3%of the total energy used in the USA.

In order to determine where technical resources should be targeted it is useful to examine the energy used in the individual processesand what the practical and theoretical limits are. Normally CO2 emissions are related to the energy used in a process. As seen inTables (1-2) hot metal, or ironmaking, are by far the largest energy consumer and CO2 emitter. Therefore, ironmaking has the greatestpotential for improvement. EAF steelmaking has improved significantly and is close to its practical minimum. In steelmaking, themajor recoverable energy is the energy in the off gas (CO). For hot rolling, starting with section sizes such as thin slab or stripreduces energy and CO2. However, substantial savings with conventional slab are possible by direct charging since reheating is themajor energy consuming part of the process. In fact, 75% of the potential saving of thin or even strip casting can be achieved withconventional slabs if they are directly charged.

Table 1. The actual, practical minimum and theoretical minimum energy consumed in various processes.

Energy (GJ/t Product)

Process Actual Practical Minimum Theoretical MinimumHot Metal 13-14 10.4 9.8

Steelmaking (BOF) 0 (0.5) (1.0)Steelmaking (EAF) 2.1-2.4 1.6 1.3

Hot Rolling 2.0-2.4 0.09 0.03Cold Rolling 1.0-1.4 0.2 0.02

( ) Indicate energy available

Table 2. CO2 emissions for various steelmaking processes

CO2 Emissions (kg/t)

Process Typical Practical Minimum Theoretical MinimumHot Metal 1500 1160 1090

Steelmaking (BOF) 200 144 144Steelmaking (EAF) 380 280 225

Rolling 320 60 10

There has been a dramatic reduction in energy consumption in the past 30 years; from about 35.8 to 18.0 GJ/tonne of shipped steel inthe US. This has been accomplished primarily due to the implementation of continuous casting, other yield improvements, a shift toscrap based EAF production and a decrease in the blast furnace fuel rate. Future reduction will be much more difficult since most ofthe “low hanging fruit has been picked”.

Strategy for Research on Sustainability

Since CO2 abatement and energy reduction are closely related and the major challenge, I will concentrate on this issue. Otheremissions such as SO2 and possible particulate matter will also be generally reduced if energy consumption is reduced.

There are at least four possible alternatives for lowering the amount of CO2; reducing the amount of CO2 emitted in the processes,using renewable energy sources such as wood charcoal, using hydrogen, and sequestering or capturing the CO2. Hydrogen will not beavailable in large quantities at a reasonable price. The available hydrogen will be primarily used for other technologies such as fuelcells and transportation. Sequestering the CO2 is a reasonable partial solution for many industries including power generation. Thesteel industry should not use “its” meager resources on this solution but rely on development from other industries. However, theindustry should examine processes which are consistent with sequestering technologies.

For example, sequestering CO2 for steelmaking may be simply accomplished by injecting CO2 into deep wells. This requires, as wellas other sequestering techniques, a relatively pure CO2 stream. Research on a oxygen blast furnace with a hydrogen shift reactor andgas circulation can produce a nearly pure CO2 off gas. CIRCOFER, which uses coal and ore in a fluid bed type process, canaccomplish the same goal. The resulting gas is nearly pure CO2 which is more easily sequestered. Research in these areas should beconsidered.

In the iron and steelmaking processes, one strategy is continuous incremental improvements with modest gains but with the total beingsignificant (10-15%). Most likely this new technology will be ironmaking because 75% of the energy is consumed in making iron.The new process should eliminate cokemaking and use fine ores, and in special circumstances, a renewable energy source.

Which strategy should be adopted? Both. We need to use the capital invested in our existing technologies while reducing energy inthese processes in the short term, 3 to 5 years. However, for the longer term 10 to 20 years, a new technology, or technologies, maybe necessary.

Research in Existing Processes

In this section and the next, research being conducted at the Center for Iron and Steelmaking Research (CISR) at Carnegie MellonUniversity aimed at reducing energy and CO2 emissions is discussed. There is extensive excellent research being done worldwideincluding Europe, Japan and other institutions in North America. I am concentrating on the CISR research program because I havemore personal knowledge.

Ironmaking: Two recent research programs in ironmaking designed to make incremental improvements in the processes are:Controlling the Cohesive Zone in the Blast Furnace and Understanding the Final Stages of Reduction by Gases.

The cohesive zone begins when the ferrous materials begin to soften and ends with final melt down. The cohesive zone impedesproper gas flow through the furnace and decreases indirect reduction by CO and H2. The cohesive zone should be deep in the furnaceand as short as possible. The cohesive zone can be improved by custom designing the ferrous burden and the charging pattern.Pellets, sinter and ore are often subjected to an arbitrary softening and melting test which give some guidance as to how they behave inthe blast furnace but do not reflect the fundamental behavior which is the objective of the present research.

Figure 1. SEM pictures from quenched samples from a laser confocal microscope

The recent work(1) included detailed characterization of the ferrous materials, melting on-set experiments using a laser confocalmicroscope, softening and melting with x-ray observations, quenched samples and thermochemical modeling of phase diagrams topredict liquid slag formation. As seen in Figure (1) it is possible to see slag formation in this acid pellet. However, the initial amountof liquid may be small and due to a minor constituent such as Na2O. More importantly is the temperature when large amounts ofliquid form and the viscosity of the liquid. This will be discussed later in detail. The lower its viscosity, the greater the impact theliquid slag has on the behavior of the pellets.

Figure 2. Experimental set-up for displacement versus temperature measurementson x-ray picture

Figure 3. Displacement versus temperature curves for acid and basic pellets.

Experiments were run in which six prereduced pellets were heated under load with the displacement measured along with thetemperature (Figure (2)). Typical displacement versus temperature curves are shown in Figure (3). In addition, x-ray pictures of thepellets were taken; examples are show in Figure (4). As can be seen, the acid pellet at 1201°C has deformed significantly with adisplacement of about 63%. The basic pellet reached 1310°C for a similar displacement and there was a melt down temperature rangewas only about 40°C. In separate experiments, the samples were quenched and the phases present determined. Calculations usingFactSage were conducted to determine the phases present. For example, for 60 prereduction of a basic pellet the phases are shown in

Figure (5). A small amount of liquid forms at about 1230°C, but a significant amount of liquid forms at 1300°C which corresponds tothe observations.

Another interesting observation is that, with the acid pellets, slag exuded from the pellet interacting with adjacent pellets, this did notoccur with basic pellets. It was concluded that this is because the viscosity of the slag in the acid pellets was much lower than in thebasic pellet. The reason the viscosity was much lower is due to the higher FeO content of the slag, about 60% versus less than 30%.

Time (s) 0 23098 24357 25237 25479 30837T (oC) 279 1141 1201 1236 1254 1502D (%) 0 45.3 63.4 75.3 80.1 100

Time (s) 0 22526 25666 26328 26792 27374T (oC) 276 1162 1310 1341 1358 1392D (%) 0 20.9 60.3 69.3 82.7 100

Figure 4. X-ray pictures of acid and basic pellets during displacement versusheating experiments.

The fundamental behavior determined in this study will aid in pellet composition design and the charging pattern when using mixedburdens. Details of this work can be found in a series of recent publications by Nogueira and Fruehan(1).

The other major method of producing iron is solid state gaseous reduction or direct reduction. This iron is primarily used in an EAF.Therefore, a high degree of reduction or metallization is desirable since the reduction of the remaining FeO by carbon is endothermicresulting in higher electrical energy consumption and lower productivity in the EAF. In the direct reduction processes, in particularfluid bed process such as FINMET or CIRCORED, it is often difficult to obtain high degrees of reduction, greater than about 92%.This phenomena depends on the ore type and process. In Figure (6) the reduction of Mount Newman Concentrate (MNC) and ahematite single crystal (HSC). MNC reaches a fairly high degree of reduction where HSC levels off at about 75% reduction(2). Thisphenomena depends on the ore type and on the overall reduction process.

The ore particles were examined by SEM analysis and found that the rate slowed when a dense layer of iron forms around theunreacted FeO, as shown in Figure (7). When this occurs, gaseous diffusion is not possible and the only mechanism for furtherreduction is atomic diffusion of oxygen through the iron layer. This process is exceptionally slow because the oxygen solubility insolid iron is only a few parts per million and hence the driving force for diffusion is low.

The reasons why the dense iron forms in some cases is not completely known. One significant observation was that the rate ofreduction slowed significantly, the wustite grains were relatively large, 1-2 µm versus 3-5 µm when the rate slowed.

Figure 5. Calculation of phases present as a function of temperaturefor a 60% reduced basic pellet.

Figure 6. Rate of reduction of Mount Newman Concentrate and hematite singlecrystals at 973 K in 60% H2-40% N2

The rate mechanism is shown schematically in Figure (8). If there is no iron layer, or it is very thin, the rate or flux or oxygen removalof removal is given by:

€

JO = −WFeOS k (PH2− PH2

e ) (1)

JO = flux of oxygenWFeO = weight of FeOk = chemical rate constant

€

PH2,PH2

e = hydrogen pressure and equilibrium pressure

Figure 7. SEM picture of HSC partially reduced showing dense iron layeraround wustite grains.

When the iron layer becomes sufficiently thick the rate becomes controlled by diffusion of oxygen through the solid iron. If sphericalFeO grains are assumed for simplicity, the diffusive flux through the iron layer is given by:

€

Jo =Do1r−1ro

(2)

where,DO = diffusivity of oxygen in Fer = radius of reaction frontro = initial radius of the particle

The rate of reduction is controlled by the slower of the two mechanisms or the one whose rate equation gives the slowest rate.

Whereas not all of the parameters required in equations (1) and (2) are known, some comparisons can be made. Using publishedvalues for the parameters indicate solid state diffusion becomes important at about 1 to 2 µm.

Obviously the grain size alone is an over simplification for the slowing of the rate for some ores and not others. However, theinfluence of grain size can be illustrated by the following example. Consider two ores, A and B, for which the rate of gaseousreduction are similar, i.e., they have similar values for the rate constant k. For ore A the wustite grain size is about 2 µm and the rateslows at about 60% reduction from wustite to iron (74% from hematite). The value of r at this point, where solid state diffusiongoverns the rate, is 1.2 µm. If for ore B the average grain size is 1 µm the value of r at which the critical rate is achieved is 0.15 µm,

which represents over 99% reduction before the rate slows. Observations indicate there is a distribution of grain sizes and not allgrains have a dense iron layer. Nevertheless, this simple example illustrates the importance of the wustite grain size on when the ratemechanism changes from chemical to diffusion control and the critical degree of reduction which can be expected.

Figure 8. Schematic diagram showing rate mechanism for solid state diffusionthrough an iron shell.

Figure 9. Experimental results indicating the effect of increasing hydrogen pressureon the rate of reduction of HSC at 973°C.

FeO

FeO

Fe

Fe

€

H2 → 2H

€

2H +O→ H2O€

H2

O

€

H2 → 2H

€

H2

H

€

H2 + FeO→ H2O+ FePH2

+ PH2O >1

Another interesting finding is when the equilibrium pressure of (H2 + H2O) at the Fe-FeO interface exceeds one atmosphere the ironshell can fracture allowing for gaseous diffusion. In this case hydrogen atoms can diffuse through the fractured iron layer and reactwith the FeO. If the total pressure is less than one atmosphere the gas cannot escape and the reaction does not proceed. As shown inFigure (9), with a hydrogen pressure of 0.6 atm reduction virtually stopped at 70%. When the hydrogen pressure was increased to 1.0atm the equilibrium pressure at the Fe-FeO interface exceeds 1 atm and presumably cracking the iron shell and allowing reduction toproceed. Also, holding or annealing the samples at the magnetite (Fe3O4) for a period of time reduction was lowered for some ores asshown in Figure (10). Presumably this may cause the grains to coarsen resulting in larger more difficult to reduce wustite grains. Theresults of this research has led to improved processing to achieve higher degrees of reduction.

Figure 10. The effect of holding reduction at Fe3O4 for 120 minutes(degree of reaction is from Fe3O4)

Oxygen Steelmaking

Research has been conducted on the recycling of waste oxides and post combustion which will be discussed in this lecture. Postcombustion (PC) is the combustion of the CO produced by decarburization to CO2 above the liquid metal bath, releasing energy. Inthe BOF, the energy can be used to melt more scrap, by about 10% of the charge; or melt skulls. However, if post combustion is notdone properly, the CO2 can react with the Fe-C melt (depostcombustion) or the heat may be released too high in the furnace and nottransferred to the bath just heating the gas or refractories.

Computational fluid dynamics (CFD) modeling was conducted to optimize the process(3). The model included an accurate combustionreaction, all relevant forms of heat transfer and depostcombustion reactions. The BOF lance had 16 post combustion nozzles. Theproblem could be simplified since there are axial systems so the model was a 3-D model of 1/16 of the furnace. Examples of thecalculated results are shown in Figures (11) and (12) which show the CO2 concentration in the gas and the temperature respectively.

This was for

€

750Nm3

s of O 2 with 18% of the oxygen used in 10 nozzles 1.5 m above the bath. This study demonstrated that CFD

could be a useful tool in optimizing post combustion and utilizing the energy associated with the CO effectively. Different cases withrespect to oxygen distribution between the primary and post combustion nozzles and height of the PC nozzles were investigated tofind the optimum condition.

Figure 11. The CO2 concentration with post combustion in a BOF

Figure 12. The temperature profile in a BOF with post combustion

Electrical Furnace Steelmaking

The steel industry can reduce energy consumption and CO2 generation by expanding EAF production to high quality steels replacingintegrated production. This will be limited due to limited quantities of high quality scrap. Processes are being developed andimproved to produce scrap substitutes such as DRI, HBI, pig iron, etc. However, these processes consume as much energy as the blastfurnace.

For the existing EAF process two examples of research aimed at reducing energy consumption are improved slag foaming and postcombustion.

Slag foaming is widely used in the production of carbon steels and extensive research on the fundamentals has been conducted. Slagfoaming allows for long arc operation during flat bath operating conditions by using the foam to shield the sidewall refractories fromthe electrical arc. However, for stainless steels, slag foaming is much more difficult to achieve. This could be due to two reasons.The slag has poor foaming characteristics or sufficient gas is not generated to create the foam. The ability of a slag to foam is givenby the foam index which is only a function of the slag properties and defined by (3).

€

Hf = ΣQA

(3)

Hf = Foam heightQ = Gas flow rateA = Area of foam regionΣ = Foam index

The foam index in laboratory experiments was found to be similar to that for carbon steelmaking slags as long as there was not anexcessive amount of solid undissolved slag particles. Typical results are shown in Figure (13) in which the foam height is plottedversus the superficial gas velocity or (Q/A) the slope is the foam index (Σ).

Figure 13. Determination of the foam index for stainless steelmaking slags byplotting the foam height versus the superficial gas velocity

The foam index for stainless steelmaking slags is similar to that for carbon steels. Therefore, the foamability or foam index is not thereason for poor foaming. Another possibility for the poor foaming is the lack of gas generation. For foamy slag practices CO gas isnormally generated by injection of carbon into the slag producing CO by the reaction:

€

(FeO) + C = Fe + CO (4)

However, for stainless steelmaking slag there is little FeO and the most reducible oxide is chromic oxide (“CrO”). The rate of reactionof carbon with slags containing FeO and CrO in which the amount of CO generated was measured in laboratory experiments. Typicalresults for FeO containing slags are shown in Figure (14). For CrO containing slags the rate was extremely slow, less than 10% of therate for similar amounts of FeO.

Figure 14. Rate of reduction of FeO in EAF slags by carbon

It therefore can be concluded that the lack of foaming in stainless steelmaking is due to poor gas generation. Consequently, foamingshould be able to be improved by making additions which generate gas. A number of additions were examined(5) such as waste oxidebriquettes, nickel oxide, calcium nitrate and limestone (CaCO3). For example, when CaCO3 is added to steel which promoted foamingit dissociates at about 900°C:

€

CaCO3 = CaO+ CO2 (5)

Typical experimental results are shown in Figure (15). It was found that the rate was controlled by heat transfer to the CaCO3 particle.Based on these results the amount and rate of CaCO3 addition was calculated to give the desired foam level.

Figure 15. Rate of gas evolution from CaCO3 in slag

Post combustion in the EAF was examined using a similar CFD model that was used for the BOF. The model was more complex dueto the lack of symmetry and the presence of scrap. The model was used to examine a system with gas injections in the sidewall of thefurnace. Typical results are shown in Figures (16) and (17) in terms of the CO-CO2-O2. As seen in Figure (16) the O2 is consumedproducing CO2 at the metal surface. CO2 re-reacts with carbon in the metal, “depostcombustion”. In Figure (17) the gas profiles andvelocities are shown. In this case, CO2 is reacting with the electrodes. The model also computes the energy transferred to the scrapand bath and electrical energy saved. The CFD model can be used to optimize the process with regards to post combustion injector’sposition and angle and the O2 flow rate.

Figure 16. Gas composition profiles for post combustion in an EAF

Breakthrough Technology

The international steel industry has been searching for an alternative to the blast furnace for over 50 years. The blast furnacecontinues to be the dominant ironmaking process because its efficiency has continuously improved and no alternative, with theexception of the direct gaseous reduction process, in unique situations, has proven to be economically competitive. However, due tothe current need for a lower capital cost process and the pressure for CO2 abatement new processes are still required. It has becomeevident that no single process will fit all needs but different processes will be used for different requirements.

Any new process should be in ironmaking where 75% of the energy and most of the CO2 is emitted. The process should use fines andeliminate cokemaking. If possible the process should use renewable energy. In addition the process should require less capital and beeconomical on a small scale.

New processes to produce liquid hot metal, which are being commercialized include: COREX, HIsmelt and processes using ore andcoal pellets or mixtures in a rotary and multi-hearth furnaces followed by electric furnace melting. Examples of the later processesinclude: Iron Dynamics, Redsmelt and Sidcomet. Whereas these processes appear attractive for several applications they do notreduce CO2 emissions. Obviously natural gas based processes such as Midrex, HyL, and FINMET reduce CO2 but natural gasavailability is limited.

Figure 17. Gas profile and velocities for post combustion in an EAF

Hot metal

OxygenHigh energyoff gas

75-80% Pre-reduced hot iron pellets

Composite pellets

Figure 18. Schematic diagram of renewal energy steelmaking process using woodcharcoal with a rotary hearth furnace and a smelter

Research is being carried out on two ironmaking process concepts which could reduce CO2 emissions. The first is a renewable energyprocess(6) using ore-wood charcoal pellets in a rotary hearth furnace which are reduced to about 70-80%. The prereduced mixture isthen added to a coal based smelter, such as an AISI or HIsmelt, for final reduction and gangue separation by melting. The smelter isrun at a conservative 30% post combustion with the off gas used in place of natural gas in the rotary hearth. A schematic diagram ofthe process is shown in Figure (18). By using wood charcoal the net CO2 emissions are reduced by 70-95% due to the fixation of CO2

when re-growing the wood. By combining the RHF with the smelter, productivity can be significantly increased (30-50%) andenergy, decreased.

Extensive research has been carried out on the reduction of charcoal pellets. Reduction takes place by means of two reactions,reduction of iron oxide (“FeO”) by CO followed by oxidation of carbon by CO2.

0

0.02

0.04

0.06

0.08

0 100 200 300 400 500

Time (min)

Ma

ss

lo

ss

(g

)

Kc = 5.0x10-4 mol/mol.s.atm

KFeO = 4.0x10-3 mol/mol.s.atm

Mixed reaction model

Experiment

(a)

0.8081g sample

0

0.02

0.04

0.06

0.08

0.1

0 100 200 300 400

Time(min)

Mass lo

ss (

g)

Kc = 5.0x10-4 mol/mol.s.atm

KFeO = 4.0x10-3 mol/mol.s.atmMixed reaction model

Experiment

(b)

0.4745g sample

Figure 19. Determination of the rate constants for reduction of FeO by CO andoxidation of carbon by CO2 in FeO-charcoal mixtures

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 2 4 6 8 10

Time (min)

Ma

ss

lo

ss

(g

)

Temperature: 1280 oC

Large Pellet Mass: 3.8599g Initial Diameter: 16.7mm Final Diameter: 12.4mm Wood charcoal/hematite: 0.1766

Simulation with final radius

Simulation with initial radius

Experiment

Figure 20. Experimental and model results for the reduction of ore-wood charcoal pellets

Figure 21. Production rate for varying pellet layer as a function of degree of reduction

€

"FeO"+CO = Fe + CO2 (6)

€

CO2 +C = 2CO (7)

The fundamental rate constants for these reactions were determined using small samples when heat and mass transfer can beneglected. Typical results are shown in Figure (19). Using these constants, and independently determined heat transfer coefficients, areduction model for pellets was developed and verified as shown for typical results in Figure (20). The pellet model was thenincorporated into a three stage model for the RHF. From this model, the production rate, as a function of desired reduction and pelletdepth was determined, typical results are shown in Figure (21). Finally the RHF model was linked to an energy and materials balance

model for the smelter to determine the overall performance of the process. The overall process has reasonable productivity and energyconsumption and greatly reduces net CO2 emissions.

Another method to reduce CO2 emissions is based on using hydrogen. However, a critical question is where to get the hydrogen from.Natural gas derives about half its energy value from hydrogen but its availability is limited. Many coals have high hydrogen orvolatile contents which cannot be normally used for cokemaking. A new process concept based on using high volatile coals which arecharred at a lower temperature is currently being investigated. A schematic of the process is shown in Figure (22). The char is used ina HIsmelt unit and the hydrogen rich off gas can be used for direct reduction or other uses, such as fuel cells. Smelters workextremely well using char in that there is much higher post combustion and productivity. For example, HIsmelt has achieved over70% post combustion using char.

Figure 22. Schematic diagram of a conceptual process to use high volatile coals to produce hydrogen and liquid iron

Concluding Remarks

The steel industry is a major consumer of energy and producer of CO2. Work must continue to incrementally lower energyand CO2 in existing processes. In a parallel and equally important effort a new breakthrough ironmaking process should be exploredand process models developed. The use of renewable energy and high hydrogen containing coals should be encouraged. Steelcompanies alone do not have the resources to meet these goals. We need international collaboration and government assistance.Finally there should be incentives for lowering CO2 not just penalties, such as carbon taxes.

Acknowledgements

The author wishes to thank the many students and researchers who conducted this research and the member companies of theCenter for Iron and Steelmaking Research, the American Iron and Steel Institute and the Department of Energy for their support.

References1. P. Nogueira and R. J. Fruehan, to be published in Metallurgical and Materials Transactions, 2004.2. R. J. Fruehan, Y. Li, L. Brabie and E. J. Kim, accepted for publication in Scandinavian Journal of Metals3. Y. Li, R. Matway and R. J. Fruehan, DOE Report, April 2004.4. J. J. Kerr and R. J. Fruehan, Metallurgical and Materials Transactions, Volume 35B, 2004.5. J. J. Kerr and R. J. Fruehan, 2000 Process Technology Conference Proceedings, ISS, pp. 1049-1063.6. O. Fortini and R. J. Fruehan, submitted for publication in Metallurgical and Materials Transactions